We surveyed adjacent reaches with differing riparian vegetation to explain why channels with forested banks are wider than channels with nonforested banks. Cross sections and geomorphic mapping demonstrate that erosion occurs at cutbanks in curving reaches, while deposition is localized on active floodplains on the insides of bends. Our data indicate that rates of deposition and lateral migration are both higher in nonforested reaches than in forested reaches. Two dimensionless parameters, α and E, explain our observations. α represents the influence of grassy vegetation on rates of active floodplain deposition; it is 5 times higher in nonforested reaches than in forested reaches. E is proportional to rates of cutbank migration; it is 3 times higher in nonforested reaches than in forested reaches. Differences in width between forested and nonforested reaches are proportional to E/α. In forested reaches, channels are wide with banks that are difficult to erode. Dense tree roots create a low value of E, and the channel migrates slowly. E/α is high, however, because α is very low: shade from trees inhibits the growth of grass on active floodplains. In nonforested reaches, channels are narrow with banks that are easy to erode. E is high, and the channel migrates rapidly. E/α is low, however, due to a very large value of α: grass grows readily on nonforested convex bank floodplains. Thus, differences in width between forested and nonforested reaches are related to a balance between rates of cutbank erosion and rates of deposition on active floodplains, implying that equilibrium widths develop to equalize rates of cutbank erosion and vegetation-mediated rates of deposition on active flood-plains. These results suggest that accurate models of width adjustment should consider the combined effects of bank erodibility and floodplain depositional processes, rather than focusing on these processes in isolation from one another.